staff blogs

distributed.net staff keep (relatively) up-to-date logs of their activities in .plan files. These were traditionally available via finger, but we've put them on the web for easier consumption.

2019-08-23

August 2019 update

Filed under: project status — Tags: , @ 02:44 +00:00

Dear friends,

Great news – the OGR-28 project has now been progressing for 2000 days and has brought 60,000 people from around the world together working towards a single goal. If you have recently joined us, welcome :)

The combined effort of our group using commodity computer equipment is producing the same amount of work that would be produced by a major supercomputer – such as the Cray XC40 model used in Texas Advanced Computer Center’s Lonestar 5. A single Cray XC40 cabinet contains 192 nodes, each with 24 cores of Intel Xeon CPU power and costs upwards of $500,000; with a power cost of about $100,000 per year.

We know that some of you have been with us since the beginning, and we are grateful for your contributions. We’ve come such a long way since 1997. A Pentium 100 computer could calculate 100,000 keys/sec of RC5. A $150 graphics card today is 30,000 times faster. Our first OGR project searching for the optimal Golomb ruler with 24 marks took a little over 4 years to complete. At the time we started OGR-24, OGR-28 looked almost impossible. Thanks to the sterling efforts of our brilliant coders, and you, our volunteers, we have proved optimal rulers for 25, 26 and 27 marks. We’re well on our way to proving the optimal ruler with 28 marks, where our throughput is 100 times faster than when we started on OGR-24. Even as we approach 70% completion on the project, there is still a better than 1 in 7 chance that the OGR stub your computer works on today has not been calculated before.

One way you can help is to download the latest version of the client software for your platform. We will be releasing new clients shortly which can support the latest processors and graphics cards from Intel, AMD and ARM. Whether you have a Core, Ryzen, Pi, Galaxy or Xeon (including Xeon Phi), we have a client for you.

Following on from her excellent work on RC5 ARM64 cores, our good friend Sunset has written a core for processors featuring AVX-512 instructions which is an astounding 70% quicker than the cores we have, which are already highly optimised.

Telling your friends about distributed.net and our projects – and working together with them in a team – will help us take months and years off our expected project completion dates. Perhaps you have a friend who used to run dnetc in our early days? Let them know that we are still here :)

Moo!